Prediction of particulate matter concentrations using LSTM networks model

Particulate matter (PM) is a major contributor to air pollution, and exposure to its harmful effects at elevated levels poses a significant risk to human health. Therefore, it is of considerable practical importance for air quality monitoring, air pollution restoration, and human health to have an accurate prediction of PM concentration. Traditional machine learning, neural networks, and deep learning are all areas of focus. In this investigation, we have undertaken a time series analysis on PM2.5 from 2018-2022 at DTU station of Delhi. For doing the time series forecasting, we have used a model called Long Short-Term Memory networks (LSTMs). Mean squared error (MAE), root mean squared error (RMSE), Mean absolute percentage error (MAPE), and Pearson Coefficient (r) are employed to check the validity and applicability of the constructed LSTM model. MSE was obtained as 1832.06, RMSE as 42.80, MAPE as 34.07 and R2 as 0.514. We have predicted PM2.5 values for next 30 days in the year 2023 based on simulated model. The values obtained were ranging from 282.6 µg/m3 to 130.9 µg/m3.